Every time you turn around it seems like some developer somewhere is struggling with latency issues.

Among the chief culprits is this massive wave of server consolidation. Developers quickly discover that applications that use to run on a local server don't perform nearly as well when they are being accessed over a wide area network. Frequently exacerbating that situation is a layer of virtual machine software that adds additional overhead to the overall performance equation.

Of course, there's also a lot more reliance on Web services protocols. While Web services are a boon to standardization, they also challenge developers because as a class of software Web services are very susceptible to latency issues.

And now with that advent of cloud computing and software-as-a-service applications, we're asking developers to create applications for an unknown number of users. Increasingly, we're trying to expose more business processes on the Web. But we can't really make any assumptions about how many simultaneous users of that service there might actually be.

All of these issues are leading to a lot of recrimination between developers and network infrastructure people. The developers are angry because they feel that the performance of their applications is being adversely affected by substandard infrastructure architecture. Conversely, the network infrastructure people say developers are creating unreasonably bloated applications that rely too much on the infrastructure to make up for their lack of programming finesse.

Stepping into the middle of this argument is Sonoa Systems, which has created an appliance that first inspects the first few elements of an XML message and then based on that information accelerates that message the appropriate application server.

On the face of it, it's a pretty simple concept that borrows principles generally associated with networking and applies them to application performance management. In doing so, it then gives developers an edge in the battle against latency in a way that doesn't require networking specialists to invest in a massive network upgrade.

Furthermore, it also provides a portal through which IT folks can actually see what is happening with the performance of a particular application, which ultimately might lead to some service level agreements that both parties can not only agree to but actually enforce by guaranteeing priority to certain types of applications.

The concept of an application accelerator has been around for while. But the Sonoa approach is a little more granular from the point of view of the developer because it takes a declarative approach to prioritizing traffic that doesn't require anyone to write additional code or recompile any software.

None of this means that application middleware is about to move wholesale into the networking layer. But it does means that there are things that can be done at the network layer to boost the performance of latency-sensitive applications. And if that can be done in a relatively transparent way, then a lot of the arguing between developers and IT infrastructure people may suddenly become moot.